Algorithmic Stablecoins Explained: How They Work and Why They Fail
By Jorge Rodriguez — Stablecoins
The burn/mint mechanics behind algorithmic stablecoins: how seigniorage loops are designed to maintain the peg
A step-by-step breakdown of the UST/LUNA collapse and why the reflexivity trap made it structurally inevitable
Seven red flags for identifying algorithmic stablecoin risk before you deploy capital
Introduction
In May 2022, the Terra blockchain's algorithmic stablecoin UST went from an $18 billion ecosystem to near-zero in less than a week. The term "algorithmic stablecoins explained" became one of the most-searched phrases in crypto almost overnight. Tens of thousands of DeFi investors lost money they had been told was sitting in a stable, dollar-pegged asset. But the collapse was not a random failure. The mechanism design of algorithmic stablecoins contains a structural fragility that becomes catastrophic under specific market conditions. Understanding it is not just historical curiosity. It is essential knowledge for anyone navigating DeFi today. This article breaks down what algorithmic stablecoins are, how the burn/mint arbitrage loop is designed to maintain the peg, why the reflexivity trap made the UST/LUNA collapse structurally inevitable, and how to recognize the red flags in any stablecoin before committing capital. For context on where algorithmic stablecoins fit in the broader landscape, see the [stablecoin risk spectrum](/blog/stablecoins/stablecoin-risk-tiers).
What Makes a Stablecoin Algorithmic?
The stablecoin spectrum runs from fully collateralized to fully algorithmic. Understanding where a given stablecoin sits on that spectrum is the first step to assessing its risk. **Fiat-backed stablecoins** like USDC and USDT hold actual dollars or dollar-equivalent reserves for every token in circulation. If you hold 1 USDC, there is one dollar sitting in a bank account or Treasury bill that backs it. The peg holds because direct redemption is always available at par. **Crypto-overcollateralized stablecoins** like DAI require more than $1 of collateral to mint $1 of stablecoin. If ETH backs a DAI position at 150% collateral ratio, the system can withstand a significant drop in ETH price before the collateral becomes insufficient. The peg holds because excess collateral can absorb volatility. **Algorithmic stablecoins** use neither of these mechanisms. Instead, they rely on code-enforced incentives and market arbitrage to maintain the peg. No external collateral is required. The stability comes from the protocol's ability to make arbitrage profitable whenever the price drifts from $1. Within the algorithmic category there are two main variants: • **Pure algorithmic:** The peg is maintained entirely by mint/burn mechanics and market incentives, with no collateral backing. UST, Basis, and Iron Finance fall here. The governance token (LUNA for UST) serves as the shock absorber for all volatility. • **Fractional algorithmic (hybrid):** A partial collateral buffer is combined with algorithmic stabilization. FRAX v1 was the primary example. A portion of each FRAX was backed by USDC; the rest was backed by the FXS governance token. The appeal of algorithmic stablecoins is **capital efficiency**. Overcollateralized systems tie up more dollars than they issue. Fiat-backed systems require centralized custodians. An algorithmic system promises a dollar-equivalent asset without either constraint. That promise attracted tens of billions of dollars before May 2022. The catch is that the stabilization mechanism depends on continuous market confidence. In normal conditions, arbitrage self-corrects the peg. Under stress, the same mechanism can amplify the depeg instead of absorbing it. That distinction between correction and amplification comes down to the structural design of the collateral.
How Algorithmic Stablecoins Maintain Their Peg
The core mechanism is a **seigniorage model** built around a paired asset system. In UST's case, this was UST (the stablecoin) and LUNA (the governance and collateral token). The protocol enforced a simple rule: 1 UST can always be minted by burning $1 worth of LUNA, and 1 UST can always be burned to redeem $1 worth of LUNA. This rule creates an arbitrage loop that theoretically keeps UST near $1 at all times. **When UST trades above $1:** If UST is trading at $1.02, an arbitrageur can burn $1 of LUNA to mint 1 UST, sell that UST in the market for $1.02, and pocket a $0.02 profit. The selling pressure from this activity pushes the UST price back down toward $1. More UST minted means more LUNA burned, which reduces LUNA supply. **When UST trades below $1:** If UST is trading at $0.98, an arbitrageur can buy 1 UST for $0.98 in the market, burn it to receive $1 worth of LUNA, and net a $0.02 profit. The buying pressure pushes the UST price back up toward $1. More UST burned means more LUNA minted, which increases LUNA supply. In theory, this is self-correcting. Rational arbitrageurs will exploit any price deviation until it closes. The protocol never needs to hold external reserves because the LUNA token absorbs the volatility.  **Rebase models** offer a different approach. Ampleforth (AMPL) adjusts the total token supply in every wallet instead of using a burn/mint mechanism. If AMPL trades above $1, every wallet holder receives more AMPL tokens. If AMPL trades below $1, every wallet holder's AMPL balance shrinks proportionally. The idea is elastic supply rather than stable price, which means AMPL is more accurately described as an elastic supply asset than a true stablecoin. **The demand driver problem:** The UST/LUNA mechanism had an additional structural vulnerability. Anchor Protocol offered 20% APY on UST deposits. This was not organic yield from lending demand. It was subsidized yield paid from a reserve fund, designed to attract deposits and drive UST demand. By early 2022, over 80% of all circulating UST was locked in Anchor Protocol chasing that yield. This concentration meant that UST demand was almost entirely dependent on one product offering unsustainable returns. When rational actors began questioning the sustainability of that yield and the Anchor reserve fund started visibly depleting, the conditions for a bank run were already in place. Understanding [how Anchor Protocol generated yield](/blog/stablecoins/how-stablecoins-earn-interest) and why that yield could not be sustained from organic protocol revenue is central to understanding why the collapse was predictable in retrospect.
The UST/LUNA Case Study: A Step-by-Step Collapse
**The Setup** By early May 2022, UST had grown to an $18 billion market cap. LUNA was valued at over $30 billion. The Terra ecosystem was one of the largest in DeFi by total value locked, and Anchor Protocol was driving massive inflows with its 20% APY on UST deposits. The Luna Foundation Guard (LFG) had accumulated approximately $3.5 billion in Bitcoin as a reserve fund to defend the peg. **The Attack Vector** On May 7, large withdrawals began from Curve Finance's UST-3pool, which held roughly $700 million in liquidity. The sell orders created a pool imbalance that pushed UST below $1. The attack was not technically sophisticated. It was simply large enough relative to available liquidity. Once the depeg began, it triggered cascading responses. **Day by Day: May 7-13, 2022**  May 7-8: UST dips to $0.985. LFG begins deploying BTC reserves, lending them to market makers to buy UST. Anchor withdrawals begin accelerating as holders watch the depeg nervously. May 9: UST drops to $0.91. The protocol begins minting LUNA to defend the peg. LUNA's price starts falling as the market absorbs increasing supply. Anchor sees withdrawal requests sharply accelerate. May 10-11: UST is trading near $0.60. LUNA hyperinflation begins. The protocol mints billions of new LUNA tokens to absorb UST selling pressure. LUNA's supply goes from roughly 350 million tokens to over 6 billion tokens within 48 hours. The dilution means each LUNA token is worth less, which means the protocol has less purchasing power to defend UST on the next iteration. May 12-13: UST falls to $0.10. LUNA is effectively worthless. The Terra blockchain is halted twice by validators trying to prevent governance attacks. By May 13, UST is trading near $0.02. The $18 billion stablecoin no longer functions as an asset. **Why the Arbitrage Loop Failed** The burn/mint arbitrage loop was designed to self-correct small deviations. It was not designed to absorb a large, rapid withdrawal when LUNA's market cap was already falling faster than the protocol could generate UST buyback purchasing power. The structural failure was circular. LUNA had value primarily because UST had demand. UST had demand primarily because Anchor was paying 20% APY. When Anchor withdrawals accelerated, UST demand collapsed, which collapsed LUNA's value, which meant the protocol's collateral was worth less as the crisis deepened. Each attempt to defend the peg by minting more LUNA added supply, which further diluted LUNA, which made the next round of defense even less effective. This is the reflexivity trap: the defense mechanism depended on the very asset whose value was being destroyed by the defense attempt. A collateralized stablecoin avoids this because the collateral's value is **exogenous**, it does not depend on the stablecoin's peg. ETH backing DAI has independent value. LUNA backing UST had no independent value floor. **The Role of Anchor Protocol** A critical point often glossed over in post-mortems: the [20% APY was not sustainable](/blog/risk-management/defi-yield-risks-explained). Anchor's yield reserve was being depleted faster than lending revenue replenished it. By early 2022, the reserve fund had dropped from over $400 million to under $100 million and was on a trajectory to run out within months. Rational actors who understood this began anticipating withdrawals. The anticipation of withdrawal is itself the trigger. When a critical mass of UST holders realized that Anchor's yield was time-limited and began exiting simultaneously, the [bank run on Anchor Protocol](/blog/risk-management/defi-bank-run-explained) was structurally inevitable, not a random shock. The $3.5 billion BTC reserve could not absorb billions of dollars of simultaneous selling at the velocity and scale of what followed.
Fractional Algorithmic Stablecoins: The Hybrid Approach
FRAX is the most significant surviving experiment in the algorithmic stablecoin space, and understanding how it differs from UST is instructive. The key distinction is the presence of external collateral as a floor. **FRAX v1: The Fractional Model** FRAX launched in December 2020 with a fractional-reserve model. Each FRAX token was partially backed by USDC (exogenous collateral) and partially by FXS, FRAX's governance token (endogenous collateral). The Collateral Ratio (CR) started at 100% USDC backing and adjusted algorithmically based on market demand. If the market had high confidence in FRAX, the CR would decrease, meaning less USDC was required per FRAX. If confidence dropped, the CR increased, requiring more USDC per token. This dynamic adjustment is what gave FRAX its fractional-algorithmic label. The partial USDC backing was the critical difference from pure algo designs. It meant that reflexivity was dampened. Even if FXS lost value, a portion of every FRAX was backed by real dollars. The floor existed. **Post-UST: The Shift Toward Full Collateralization** The UST collapse changed FRAX's trajectory fundamentally. The FRAX team recognized that endogenous collateral at scale was structurally dangerous and moved decisively away from the v1 architecture. FRAX v3 introduced sFRAX, a yield-bearing vault that deploys collateral into T-bill-backed real-world assets, generating real yield rather than protocol-subsidized returns. FraxBonds (FXBS) were introduced as an additional instrument within the system's capital structure. The algorithmic component that defined v1 has been progressively reduced, and the protocol now operates with collateral ratios substantially higher than the original fractional design intended. Understanding how [collateral backing changes risk profile](/blog/stablecoins/usdc-vs-usdt-defi-safety) is directly relevant here. The more a stablecoin's backing consists of assets with independent value, the less reflexive its failure mode becomes under stress. **What FRAX Does Not Solve** Surviving the UST collapse does not mean FRAX is risk-free. FXS governance token risk remains: a sustained FXS price decline creates recapitalization pressure. Oracle risk persists because FRAX relies on price feeds for collateral calculations. Smart contract complexity in v3 introduces more surface area than the original design. FRAX survived because its design was more conservative than UST and because the team adapted post-collapse. "Survived so far" is not the same as "safe." But it represents a meaningfully different risk profile from pure algorithmic designs.
The Death Spiral: How Reflexivity Kills Algorithmic Stablecoins
The abstract principle behind the UST collapse has a name in financial theory: **reflexivity**. The concept was articulated most by George Soros in the context of currency markets, but it applies directly to algorithmic stablecoin design. In a reflexive system, the price of an asset depends on beliefs about that asset's value, and those beliefs are themselves influenced by the price. A falling price causes negative beliefs, which cause further selling, which causes further price declines. The feedback loop is self-amplifying rather than self-correcting. Algorithmic stablecoins with endogenous collateral are maximally reflexive. Here is the mechanism, step by step: • UST depegs slightly from a large sell order, external shock, or loss of confidence. • The protocol mints more LUNA to buy and burn UST, defending the peg. • The new LUNA supply dilutes existing LUNA holders, reducing LUNA's price. • With lower LUNA price, the protocol has less purchasing power to defend UST on the next cycle. • UST depegs further, triggering the same sequence again at a worse ratio. • The loop continues until either confidence restores (possible under mild conditions) or the spiral runs to zero (inevitable under sufficient stress). A collateralized stablecoin breaks this loop because the collateral's value is **exogenous**. It does not depend on the stablecoin's peg. If ETH backs DAI and ETH price drops, DAI can still be defended by liquidating the ETH into dollars. The collateral has independent value that exists regardless of whether DAI holds its peg. With UST and LUNA, LUNA had no independent value floor. LUNA's value came from UST demand. UST demand came from Anchor yield. When Anchor yield became unsustainable, the entire value structure was circular, and a circular structure can only collapse simultaneously. **The bank run analogy:** The anticipation of failure causes failure. This is exactly the dynamic in traditional bank runs. Rational actors do not wait to see if the situation stabilizes. If the expected value of holding UST drops below the expected value of exiting, they exit, and their exit makes the situation worse for those who remain. This is why [circular risk in DeFi protocols](/blog/risk-management/counterparty-risk-defi) is one of the most dangerous structural patterns in the space. Liquidity conditions determine whether the loop resolves or spirals. In normal markets, small deviations self-correct. In stressed markets with rapid exits, the same mechanism becomes catastrophic. That difference is predictable from the structural design, not a matter of bad luck.
Can Algorithmic Stablecoins Ever Work?
The honest answer: possibly, under strict constraints that most projects have not respected. But the track record is not encouraging. The academic debate predates UST. The 2018 Basis protocol whitepaper laid out a theoretically sound model for seigniorage-based stablecoin design. Iron Finance, a fractional model that collapsed in June 2021, provided an early warning that fractional and algorithmic hybrids were vulnerable to bank run dynamics at scale. UST proved the point definitively at $18 billion. **What the surviving experiments share:** • Partial or full external collateral buffer, not pure endogenous collateral • Conservative collateral ratio adjustments, not aggressive algorithmic pivots toward minimal backing • No unsustainable yield driver creating artificial demand concentration in a single product • Protocols that contracted conservatively during stress rather than expanding to maximum scale The [Bank for International Settlements working paper on stablecoin design](https://www.bis.org/publ/work1014.pdf) identifies the core tension: capital efficiency and decentralization create structural fragility. The more capital-efficient a stablecoin design, the more it relies on market confidence rather than hard collateral, and the more reflexive its failure mode becomes under stress. **The scaling problem:** Algorithmic stablecoins become more fragile as they grow. A $100 million purely algorithmic stablecoin is survivable because the governance token market cap can plausibly exceed the stablecoin market cap. A $20 billion purely algorithmic stablecoin requires a governance token market cap that dwarfs the stablecoin, a condition that becomes increasingly difficult to maintain as inflows accelerate and the governance token is simultaneously being diluted to defend the peg. **The regulatory context:** The EU's MiCA framework took effect in 2024 and [restricts algorithmic stablecoins from payment use](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32023R1114) in European markets. Under MiCA's asset-referenced token rules, stablecoins that rely on endogenous collateral face significant limitations on their use as payment instruments. This does not make holding or trading them illegal, but it removes a major pathway to mainstream adoption. The takeaway for investors: understanding [why unsustainable yield is a structural warning sign](/blog/yield-strategies/yield-sustainability-defi) matters most in the context of algo stablecoins. Any algorithmic stablecoin where demand is primarily driven by subsidized yield rather than organic utility is carrying a structural countdown timer, not a stable peg.
How to Identify Algorithmic Stablecoin Risk Before You Invest
Most UST holders missed the warning signs because 20% APY overwhelmed risk analysis. The signals were visible before May 2022. Here are seven to check before deploying capital near any stablecoin with algorithmic or hybrid characteristics.  **1. Endogenous-only collateral** If the governance token backing the stablecoin is issued by the same protocol, the system is circular. Ask: if the stablecoin loses 50% of its peg, what happens to the governance token's value? If the answer is "it also falls," you have reflexivity risk with no independent floor. **2. Unsustainable yield as the demand driver** If the primary demand for the stablecoin is a yield product that cannot be sustained from organic protocol revenue, that yield is a temporary subsidy. Model the sustainability: where is the money actually coming from, and what happens when the subsidy fund runs out? **3. Governance token market cap below stablecoin market cap** If the governance token's market cap is smaller than the stablecoin's market cap, the protocol cannot defend the full peg with its collateral. Watch this ratio over time. A narrowing gap signals increasing fragility, not a number to check once. **4. No transparent proof of reserves** If you cannot verify collateral ratios on-chain or through published attestations, assume the worst. Legitimate protocols publish verifiable on-chain data. Opacity about reserve composition is a risk flag, not a design choice. **5. Rapid TVL or supply growth** Fast growth in an algorithmic system accelerates reflexivity risk proportionally. A protocol growing 100% per month is also becoming 100% per month more dependent on continued inflows to sustain the growth. Organic, measured growth is a sign of healthier underlying demand. **6. Single liquidity pool dependency** UST's dominant liquidity was in one Curve pool. One large coordinated withdrawal from that pool was enough to initiate the cascade. Healthy stablecoins have liquidity spread across multiple venues, chains, and pool types, so no single withdrawal can break the peg. **7. Centralized peg defense mechanism** Decentralized governance sounds safe. But if the real peg defense in a crisis depends on a foundation reserve controlled by a small team (as with LFG's BTC), you have a single point of failure. Ask who actually defends the peg under stress and whether that mechanism is transparent and on-chain. The [risk tier framework for stablecoins](/blog/stablecoins/stablecoin-risk-tiers) provides a broader classification approach for evaluating any stablecoin, not just algorithmic ones. The [Lince stablecoin yield tracker](https://yields.lince.finance/tracker/solana/category/stablecoin) lets you monitor stablecoin yields and peg deviation data in real time. The data that matters before a crisis is available before the crisis. After is too late.
FAQ
### Are algorithmic stablecoins illegal? Not universally. The EU's MiCA framework restricts pure algorithmic stablecoins from being used as payment instruments in Europe under its asset-referenced token rules. In the US, regulatory guidance on algo stablecoins remains incomplete as of 2025, though enforcement actions have targeted specific projects. Holding or trading them is not illegal in most jurisdictions, but their path to mainstream payment use is increasingly restricted by regulation. ### What is the difference between UST and DAI? DAI is overcollateralized by external assets with value independent of DAI's peg -- ETH, WBTC, USDC, and real-world assets. DAI's collateral does not derive its value from DAI demand. UST was backed by LUNA, a token whose value was almost entirely dependent on UST demand. That circular dependency is the structural difference. DAI can withstand its collateral losing value because the overcollateralization provides a buffer. UST had no buffer independent of the system itself. ### Is FRAX still an algorithmic stablecoin? Not in its original form. FRAX v1 was fractional-algorithmic. Following the UST collapse, FRAX progressively increased its collateral ratio and moved toward near-full exogenous collateral backing through v2 and v3. FRAX v3 introduced sFRAX and RWA-backed yield mechanisms as part of a broader shift away from algorithmic stabilization. The current architecture is more accurately described as a hybrid or partially-collateralized stablecoin with significantly reduced algorithmic exposure compared to its 2020 design. ### Could UST have been saved? This is debated. The LFG Bitcoin reserve of approximately $3.5 billion was insufficient at the scale and speed of the collapse. Some analysts argue earlier deployment could have slowed the spiral. The structural view: once confidence broke at scale in a pure endogenous-collateral system, there is no independent floor to arrest the decline. The arbitrage loop reversed direction and accelerated the collapse instead. ### What stablecoins are actually safe? No stablecoin is risk-free. They trade different types of risk. Fiat-backed stablecoins with regulated reserves (USDC, PYUSD) carry the least structural depeg risk but have counterparty and censorship risk. Overcollateralized onchain stablecoins (DAI) carry oracle and governance risk. Delta-neutral synthetics carry funding rate risk. The right question is which specific risks you are willing to accept for a given yield level. ### What happened to other algorithmic stablecoin projects? Most failed before UST at smaller scale. Basis shut down in 2018 due to regulatory concerns before it fully launched. Empty Set Dollar and Dynamic Set Dollar collapsed in 2021 when reflexivity dynamics played out at smaller TVL levels. Iron Finance collapsed in June 2021 in a direct preview of the UST dynamic, with its TITAN governance token going to zero in hours. The pattern across all of these is consistent: rapid growth, demand concentrated in unsustainable yield, governance token market cap compression, then reflexivity spiral. UST was larger, but the mechanism was the same. ### How does peg deviation signal risk? The size of a deviation matters less than its persistence and velocity. A brief dip to $0.99 that recovers in hours suggests short-term liquidity pressure. A sustained deviation that does not recover, or one that is deepening over hours despite no obvious external shock, signals that the arbitrage mechanism is not functioning as designed. Checking on-chain liquidity pool imbalances alongside price deviation is useful -- a significantly imbalanced Curve or Uniswap pool often precedes a more significant depeg by hours, not minutes.
Conclusion
Algorithmic stablecoins are structurally fragile because of how they solve the capital efficiency problem. When stabilization depends on endogenous collateral and market confidence, the system works until it does not, and when it stops working, the failure is catastrophic rather than gradual. The UST/LUNA collapse proved this at $18 billion. Iron Finance proved it at $2 billion. Basis, Empty Set Dollar, and Dynamic Set Dollar proved it before either of those. The useful takeaway is not "avoid everything that sounds algorithmic." It is the seven-signal checklist: endogenous collateral, unsustainable yield drivers, governance token market cap relative to stablecoin market cap, reserve transparency, growth velocity, liquidity concentration, and centralized peg defense. Any one of these signals warrants caution. Multiple signals appearing together warrant exit. If you hold or are evaluating any stablecoin yield position, monitoring peg stability and collateral data before a crisis is the only time that information is actionable. The [Lince stablecoin yield tracker](https://yields.lince.finance/tracker/solana/category/stablecoin) tracks yield and peg data across stablecoins and protocols so you can see what the market is pricing in before it becomes a headline.